Ferenc Huszár (@fhuszar) 's Twitter Profile
Ferenc Huszár

@fhuszar

Secular Bayesian.
Associate Professor in Machine Learning @Cambridge_CL. Talent aficionado at airetreat.org
Alum of @Twitter, Magic Pony and @Balderton

ID: 438301250

linkhttp://inference.vc/about calendar_today16-12-2011 12:31:45

14,14K Tweet

40,40K Followers

1,1K Following

Ferenc Huszár (@fhuszar) 's Twitter Profile Photo

A simple back-of-the envelope calc: When you work on AGI, each second you spend not working on AI safety increases p(doom) by, say, 10^{-10}. Thus, it would be immoral NOT to buy a $3m megacar that holds the current record for the 0-248mph-0 sprint to do your Whole Foods runs.

David Pfau (@pfau) 's Twitter Profile Photo

Reminder that I did include formal math in the bottom left corner here. I'm very impressed by AlphaProof but I think it doesn't say much about AI "in the real world".

Petar Veličković (@petarv_93) 's Twitter Profile Photo

If you're interested in NAR or CO, check out Euan Ong's poster, recasting the algorithmic bottleneck as an **ensembling bottleneck** ++ a lightweight, nonparametric solution (pDAB) to alleviate it. Stop by the Differentiable Almost Everything Workshop #ICML2024 @ Stolz 0 today!

If you're interested in NAR or CO, check out <a href="/euan_ong/">Euan Ong</a>'s poster, recasting the algorithmic bottleneck as an **ensembling bottleneck** ++ a lightweight, nonparametric solution (pDAB) to alleviate it.

Stop by the Differentiable Almost Everything Workshop #ICML2024 @ Stolz 0 today!
Ferenc Huszár (@fhuszar) 's Twitter Profile Photo

AlphaProof result could be anticipated knowing the components of the solution. Still, it is depressing how fast and inevitable the progress is.

AlphaProof result could be anticipated knowing the components of the solution.

Still, it is depressing how fast and inevitable the progress is.
Alexander Terenin (@avt_im) 's Twitter Profile Photo

Reminder that when NeurIPS papers are spuriously rejected, it wastes everyone's time: * Authors do extra unnecessary work * A whole new set of reviewers have to write reviews To save time for us all, please read rebuttals carefully, correct misunderstandings, and score fairly.

Ferenc Huszár (@fhuszar) 's Twitter Profile Photo

If we normalise Olympic medal count by population, Hungary is #8, but… If we normalise for academic freedom index instead we might make it to the top three!

If we normalise Olympic medal count by population, Hungary is #8, but…

If we normalise for academic freedom index instead we might make it to the top three!
Ferenc Huszár (@fhuszar) 's Twitter Profile Photo

Teams of Hungary took 5th and 6th place in the first International Olympiad in AI taking the top 2 spots for silver 🥈. Congrats to the participants who brought a lot of self-taught expertise, adaptibility to the event continuing to build our tradition of problem solving. 🇭🇺🇭🇺🇭🇺

Teams of Hungary took 5th and 6th place in the first International Olympiad in AI taking the top 2 spots for silver 🥈.

Congrats to the participants who brought a lot of self-taught expertise, adaptibility to the event continuing to build our tradition of problem solving. 🇭🇺🇭🇺🇭🇺
Ferenc Huszár (@fhuszar) 's Twitter Profile Photo

With all these new Python package manager things I’m seeing: is it worth paying attention to all the gazillion competing things now or should I wait until the dust settles and there’s clear advice?

Ferenc Huszár (@fhuszar) 's Twitter Profile Photo

The new International Olympiad in AI (for high schoolers) is looking for nominations to the scientific committee. Important and impactful role, helping to shape what material/topics/problems talented teenagers first encounter in the topic. Reach out if you’d consider contributing